The Big 3 Legal Automation Problems
And why we need better interlocutors.
3 challenges in legal automation and everything stopping you from having a perfect AI agent will fall into one of these buckets.
In receding order of pain:
1. Docx formatting, review and manipulation.
2. Long context review and drafting (high word count).
3. Prototyping and testing.
Let's break it down.
1. There is no wishing away MS Word. It's an unhappy marriage between wordsmiths and innovators because it is least friendly to bringing the two together. No love but we have to make do.
2. When reading a LOT of stuff in a high stakes matter, no matter what RAG or models you use, it isn't a substitute for human review. It accelerates, it helps catch things humans have missed, and it does a better job than most humans. But you still have to trust your instinct. In high stakes matters, it isn't the % of accuracy that matters as much as WHAT you got wrong. One mistake could make your whole review pointless and a high liability slide to embarassment.
3. Speaking to people who don't understand your problems, and then having to test and provide feedback on whether something is working and endlessly, is just too much pain and if it culminates in a place that you are better off not having embarked on, well...
Our studio Axara AI has a place in the world that also has OpenAI, Google, Anthropic and Mistral because we need lawyers who translate, lawyers who know, lawyers who care and lawyers that build.
Tiding over actual problems is going to create immense value in the next few decades.
Be a part of it and soon.